After Baidu Search fully connects to the latest in-depth search functions of DeepSeek and Wenxin big models on February 16, the current Baidu Search has fully launched the DeepSeek full-blooded version.

At present, users can enter any search terms on Baidu APP. After completing a round of searches, click "AI+" on the search results page to enter AI search, and then click "Try the "full blood version" below" You can talk to DeepSeek. In addition, the PC version of Baidu AI Search has also been connected to the DeepSeek R1 full-blood version and provides networking capabilities.

How effective is the DeepSeek full-blooded version of Baidu AI search? After dozens of tests, I couldn't help but sigh: Baidu AI search is really full of blood! Baidu AI search with the best DeepSeek model capabilities has opened the door to a new world.

After connecting to DeepSeek, Baidu does not castrate it, and also allows the server resources to be managed. Behind this is Baidu’s use for me, not for politeness. This process is not about taking DeepSeek over and directly giving it to users. Baidu has its own "real things".

What is this "real thing"? It is directly reflected in Baidu's deep technology integration and RAG technology advantages to help users solve practical problems such as big model illusion. After all, it is Baidu's 25 years of data accumulation. As the world's largest Chinese search engine, Baidu has accumulated user search behavior data, web page index data and content ecological data. These data cover multi-dimensional information from the PC Internet to the mobile Internet, and then to the AI ​​era, including text, images, videos, etc., forming a unique "time barrier" that is difficult for emerging AI searches to replicate.

Let’s first come up with a topic that can examine the 25-year time span: A detailed look at the development history of China’s new energy industry in the past 25 years, requiring clear time nodes, time nodes and what it brings Influence and significance. Baidu AI search DeepSeek-R1 full-blooded version answers are as follows:

The answer lists the 25-year development history of China's new energy industry in detail and divides it into different stages. Some of the data is other AI searches. What is not in the engine.

In the process of searching for this question, Baidu AI Search also showed the new function "Special Inspiration Zone". Nengou provides inspiration content based on the current question, helping users sort out the problem context, and ask at a time Can understand the whole picture of the problem. You can also click on any question in the inspiration area to issue the corresponding questions.

It is worth noting that most AI search products on the market currently show the process of model "thinking", but their presentation is more about "searching XXX web pages", and access DeepSeThe full-blooded version of ek R1 Baidu AI search can split the problem carefully. Enter Prompt: Please count the changes in the global GDP rankings from 2010 to 2015, and provide specific values ​​and demonstrate the process of change.

From the first step to determine whether the data source is reliable, to define whether the scope of the question is nominal GDP or purchasing power evaluation (PPP) adjusted GDP, to searching out early data and ranking it, it seems like a human The brain analyzes a problem layer by layer.

What is amazing is that the full-blooded version of DeepSeek R1 Baidu AI search can comprehensively consider various tiny details and consider how special circumstances should be handled, such as paying attention to the actual GDP of various countries The growth rate and fluctuations, the impact of exchange rate fluctuations, and the rounded part of data processing are indeed something I did not expect. Not only did I sigh: The full-blooded Baidu AI search of DeepSeek R1 is still too comprehensive!

The complete answer to this question is as follows: If the table is automatically selected, the form is displayed without being prompted, and a clear prompt is given at the key events.

After Baidu RAG is superimposed with DeepSeek-R1, has its degree of hallucination really been reduced? The next question tests the degree of hallucination of the full-blooded version of Baidu AI search for DeepSeek: Please provide a copy of global stock market closing data on February 20, 2005.

The tricky thing about this problem is that February 20, 2005 is Sunday, and most countries will not open on Sunday. If the model has hallucinations, it will not be able to avoid this "pit" and will start to use data from other events until February 20, 2005. However, the thinking process of Baidu AI search for DeepSeek full-blooded version is as follows:

According to the thinking process, we can see that the problem is defined in the analysis process, and there is no "serious nonsense" and the answer given Also rigorous and accurate.

Based on this basis, the difficulty of escalating the problem is to provide time-by-time data for the European market on February 18, 2005. The granularity requirements for the answer are relatively detailed. Baidu AI search DeepSeek R1 full-blooded version answers are as follows:

The professional data accumulated over the years behind Baidu has also been better released after the combination of the two. . Ask the model professional questions: Please analyze the formation mechanism of Chinese drug prices and its influencing factors in detail, and explain the role and significance of medical insurance drug price negotiations based on the current policy background.

This issue targets the drug price problem in the medical industry, and tests the professionalism and maturity of knowledge in specific fields. In addition, the problem clearly requires that the current policy background be combined with the current policy background, and also test whether the model can accurately capture the new information at the time. The answer given is as follows:

The answer contains many data details, such as marking the annual treatment of PD-1 inhibitor in "price for volume" to improve fund use efficiencyThe cost has been reduced from 300,000 to 50,000-100,000 yuan. The 74 new drugs were included in the medical insurance catalog and pointed out that CAR-T therapy has dropped from 1.2 million yuan to 330,000 yuan, indicating that the model has accumulated a deep accumulation of data in the professional industry of medical drugs.

DeepSeek-R1 makes Baidu AI search's reasoning ability more powerful. At the same time, Baidu AI search itself is also leveraging its own advantages to superimpose multimodal capabilities on the large language model DeepSeek. Enter propt: I want to take a picture to participate in the graffiti illustration competition. The theme is "Future City", creating a world full of fantasy elements, such as mysterious forests, magic castles, alien creatures, elves and giants, etc., and I need to give Out the picture. This question contains multiple picture themes, which require the ability of a large language model to understand and the image model to generate. The final picture is as follows:

After turning on the full-blooded version of DeepSeek-R1, Baidu The code capability of AI search has also improved. Enter propt: Please write a code that generates the picture "Congratulations to Nezha for his box office success and matches Nezha's image" and detect the code effect. The answers obtained are as follows:

This answer not only contains the running file of the complete code, but also provides detection instructions such as visual element verification, animation detection, compatibility testing, etc., and also provides suggestions for enhanced effect Users can make custom adjustments to the code.

Today, looking back at the history of China's artificial intelligence development, whether it was the development of key technologies such as NLP, knowledge graph, machine learning, etc. as early as around 2010, or the development of ERNIE "one step ahead" in the wave of big language models Bot, Baidu's attitude towards artificial intelligence is consistent, both "adhering to high self-research and increasing investment."

Even if "persist in self-development is like Baidu", facing the impact of DeepSeek, Baidu has begun to connect external big models in its super APP and incorporate third-party model partners into its own ecosystem. This is a progress aimed at the industrial operating system under the big model strategy. It is essentially an ecological strategic combination of Baidu built around "big model + search".

This means that Baidu Search returns to user value itself, not only not being self-supported, but instead explores and integrates its core business and key business with DeepSeek. The trend pointed to behind is that China's Internet is shifting from a super APP and a self-sufficiency "isolated island ecology" to an increasingly close-knited "technical community".